"fin" version - trained with 15 epochs, but fewer steps, better tagging
just a model for testing, use it at your own risk 😜
